Search papers, labs, and topics across Lattice.
4
0
7
Data referencing errors plague LLMs even in structured tasks, but a lightweight critic model can boost accuracy by up to 12%.
LLMZero uncovers that adaptive training strategies can boost RL performance by up to 140% by dynamically adjusting regularization parameters in response to training dynamics.
Systematic feature engineering can outperform advanced model architectures, closing a critical evaluation gap in tabular benchmarks.
Scaling synthetic environments with automatically generated tasks and verifiers unlocks significant reasoning improvements in language models, achieving a 27% relative gain on BBEH.
